Skip to content

What Is an Agent-Hour?#

An agent-hour is the billing unit for Lens Agents. One agent using the platform for one hour equals one agent-hour.

How It's Measured#

Measurement differs by agent type:

Agent type Measurement Billing trigger
Desktop AI tools Active use time — when the tool is connected and making requests through Lens Agents Clock starts when the tool connects, stops when it disconnects or is idle
External agents Active use time — when the agent is connected and making requests Clock starts on first request, stops after a period of inactivity
Managed agents Deployed time — billed for as long as the agent exists Clock starts at creation, runs continuously until the agent is deleted

Managed agents are billed continuously because they are always available — heartbeat monitoring, scheduled tasks, and Slack responses can fire at any time. Desktop tools and external agents are billed only for active use.

What's Not Included#

LLM costs are separate. You pay your model provider directly (Anthropic or AWS Bedrock for managed agents; whichever provider your desktop or external agent is configured to use). For managed agents, LLM usage is tracked in Cost Explorer at provider rates with no markup.

Agent-hours cover the platform: governance, sandbox execution, credential injection, audit trail, connections, and administration.

Agent-hours and spending controls#

Agent-hours measure platform usage. Spending controls measure LLM spending. They are independent:

Agent-hours Spending controls
What it measures Platform usage time LLM token costs
Billed by Lens Agents Your LLM provider
Controls Plan limits and overages Budget caps (daily, weekly, monthly)
Applies to All agent types Managed agents (via LLM proxy)

Spending controls cap LLM costs; agent-hours are the platform fee. Both are visible on the platform alongside the Cost Explorer.

Free tier#

Every organization gets 100 free agent-hours per month on the Standard plan. This is enough for a single developer using desktop tools for about 5 hours per workday, or one managed agent running for about 4.5 days.

Estimating LLM costs#

LLM costs are separate from agent-hour pricing and depend on your model provider, model choice, and usage patterns. Lens Agents tracks LLM usage for managed agents through the LLM proxy but does not mark up provider costs.

Typical token usage per operation:

Operation Input tokens Output tokens Frequency
Heartbeat check 5,000–15,000 1,000–5,000 Per interval (e.g., every 2 hours)
Scheduled task 3,000–10,000 1,000–8,000 Per schedule
Chat response 2,000–20,000 500–4,000 Per message
Classification gate (alert suppression) 1,000–3,000 100–500 Per heartbeat

Token usage varies significantly based on the agent's workspace files (loaded every invocation), conversation history, and task complexity. Use the Cost Explorer to monitor actual usage after deploying agents.

Desktop and external agent LLM costs are billed directly by your model provider. These costs are not tracked in Lens Agents — only managed agent LLM usage flows through the platform's LLM proxy.